0
$\begingroup$

I've RNA seq, Human, Paired-end data, Sample size is <40. These are aligned using STAR, RSEM processed. With RSEM I've TPM and expected counts, that is two files columns as individual IDs and row as gene names.

I'm interested to normalize gene data. With edgeR tutorial (link in the end) and few other online resources I see that after following steps there's an R object that contains norm.factors (Page 15) value for each individual.

I'm unable to wrap my head around it, now. If I'm interested to get normalized gene counts, can I go ahead and multiple each individual's norm.factor into its gene counts? For example, the expected count for IND1 for Gene1 is 100 and it's norm.factor is 0.80, can I say that the normalized gene count is 100*0.80=80?

I'm not interested to perform voomm, or differential expression analysis.

Follow up question: If I'm to sample few individuals' data from these RNA-seq should I
a) normalize all together and extract interested individuals, or,
b) extract interested individuals and then perform normalization.
I think since normalization uses per person library size, each person's normalization is independent of the others. Can someone please correct me if I'm wrong?

https://www.bioconductor.org/packages/release/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf

$\endgroup$

1 Answer 1

2
$\begingroup$

If I'm interested to get normalized gene counts, can I go ahead and multiple each individual's norm.factor into its gene counts? For example, the expected count for IND1 for Gene1 is 100 and it's norm.factor is 0.80, can I say that the normalized gene count is 100*0.80=80?

No; you divide, but if you poke around, you can probably find a way to get edgeR to return the normalized counts.

I think since normalization uses per person library size, each person's normalization is independent of the others.

Not true for the default normalization used in edgeR, TMM. CPM normalization is independently determined for each sample.

In general, normalize everything together, unless you think that the algorithm assumptions might be violated by including all of them together.

$\endgroup$
2
  • $\begingroup$ I was thinking to get the quantile normalization as on the question link: bioinformatics.stackexchange.com/questions/2586/… Thank you for your kind reply. :) $\endgroup$ Sep 6, 2019 at 20:52
  • $\begingroup$ I got what I was looking for thank you @swbarnes :) poke around thing helped :D $\endgroup$ Sep 6, 2019 at 21:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.